Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Update finetune and oneshot tests #114

Merged
merged 6 commits into from
Sep 5, 2024
Merged

Update finetune and oneshot tests #114

merged 6 commits into from
Sep 5, 2024

Conversation

dsikka
Copy link
Collaborator

@dsikka dsikka commented Aug 26, 2024

SUMMARY:

  • Properly update finetune tests such that they can run on a nightly cadence. This includes defining a simple config file that can be used by the tests which defines the cadence
  • Mark finetune tests which are not integration tests as unit
  • Seems as oneshot does not work with auto - update the tests to use cuda: 0

@dsikka dsikka requested a review from Satrat August 26, 2024 20:52
Copy link
Collaborator

@kylesayrs kylesayrs left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Nice

Comment on lines +117 to 120
- name: "🔬 Running transformers tests"
if: always() && steps.install.outcome == 'success'
run: |
pytest tests/llmcompressor/transformers/compression -v
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

What about the tests under obcq/gptq/oneshot/sparsification?

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

They're all running. Lines 117 onwards. Just separating them out into separate steps

device: "auto"
device: "cuda:0"
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

What was the error you were getting with auto? I think we want at least one test confirming that auto is functional

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

"Unspecified CUDA launch errors" - this was on an a100

@dsikka dsikka requested a review from Satrat August 28, 2024 19:32
@dsikka dsikka merged commit 2135c4c into main Sep 5, 2024
6 of 7 checks passed
@dsikka dsikka deleted the swap_auto branch September 5, 2024 02:33
markmc pushed a commit to markmc/llm-compressor that referenced this pull request Nov 13, 2024
…ue with the W4A8 representation to have dynamic token for activations (vllm-project#114)
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants